-
## 一言でいうと
強化学習において環境のシミュレーター(World Model)とエージェントの操作(Controller)を分けて考えたモデル。環境は画面の表現をVAEで、時系列の遷移をRNNで学習(次時刻におけるVAEの潜在表現zの分布を予測する)、操作側はVAEの潜在表現とRNNの隠れ層を結合して重みをかけるだけというシンプルさ。
シミュレーターとコントローラーは別個に学習する…
-
The unlearning challenge could benefit from accounting for the impact of network architecture and training methods on unlearning performance.
Some neural network architectures and training methods,…
-
# Paper
Link: https://www.nature.com/articles/s41586-019-1119-1
Year: 2019
# Summary
Synthesize speech from brain activity (ECoG), using encoder-decoder network.
Technology that translates ne…
-
## 一言でいうと
テキスト分類を行う際に、単語のone-hotではなく埋め込み表現に摂動を挿入し正則化効果を得る手法。加える摂動は分類を邪魔できたかと、元のデータとどれだけ離せたかの2種で検証(前者は教師ラベルが必要だが、後者は不要になる)。教師なしの後者でも十分な効果があることを確認。また、併用も可能。
### 論文リンク
https://arxiv.org/abs/1605…
-
## 一言でいうと
2つのDNNの類似性を調べる試み。同じレイヤー内のノードについて、全入力に対する出力を計測し、出力結果をSVDにかけて固有ベクトルを抽出、それらの相関をカノニカル相関でとるという形で出力の類似性を比較する。これにより比較を行えるほか、学習前後の層を比較することで学習状況を可視化できる。
### 論文リンク
https://arxiv.org/abs/1706.…
-
Let's schedule a meetup towards the end of April:
### Speakers
- Jörg Herbel (jherbel)
- Wolfgang Giersche (smurve)
- Titus von Köller
### Date
- Tuesday, 24.04.2018
### Location
- tbd
-
Post your question here about the orienting readings: “[Preface: How to Think with Deep Learning](https://docs.google.com/document/d/1_IiBXAPDNmGHXk2ETmr85eKT9BgjOFqB72D0nRsqnEY/edit?usp=sharing)”, “[…
-
**Task** (what are you trying to do/register?)
I am trying to build a template for whole body sagittal T2 MRI from a dataset of ~1000 images. These are 2D images. Please see example below:
![image](…
-
Good afternoon, the problem is the following - in the browser, using the module, changes clothes perfectly. Unniffed the request, prescribed parameters in the api http://127.0.0.1:7860/sdapi/v1/img2im…
-
### Search before asking
- [X] I have searched the YOLOv8 [issues](https://github.com/ultralytics/ultralytics/issues) and [discussions](https://github.com/ultralytics/ultralytics/discussions) and fou…